What might be behind the reported effect, that the sonic qualities of very small value bypass capacitors are remarkably imparted on the sound?
If it's into the hundreds of kHz range where the bypass cap has its lowest impedance and is somewhere like at least tens of ohms in the audible range, how could this translate to audible improvements?
I believe it does, enough credible-seeming people say they hear the difference. Would like to understand this phenomenon, technically.
What kind of non-linear interactions might happen when using a 0.047uF or 0.1uF capacitor in parallel to, say, some tens of uF?
If it's into the hundreds of kHz range where the bypass cap has its lowest impedance and is somewhere like at least tens of ohms in the audible range, how could this translate to audible improvements?
I believe it does, enough credible-seeming people say they hear the difference. Would like to understand this phenomenon, technically.
What kind of non-linear interactions might happen when using a 0.047uF or 0.1uF capacitor in parallel to, say, some tens of uF?
Let's hope this discussion will concentrate on the "science" as stated in the thread title!
Yes, let's hope there is some science to be found on this.
I'm sure there is, but it might be some piquant research that's kind of... compartmentalized. Then again the workings of (bypass) capacitors is very well known already, definitely not by me, but I know it's known. Applying all that knowledge on using small value bypass film/foil caps in a crossover seems to often result in disagreement, however... As many things in "audio science", trying to fit and match known nature's law to a phenomenon that's as complex as titillating thousands of microscopic hairs that feed into a complex chain of gain and feedback processors isn't the simplest of things. I personally appreciate the input of many knowledgeable people who have wrote informative posts that essentially are against the notion that bypass caps would be useful - however, I also am not ready to believe that so, so many people describing their enjoyment from bypass caps are hearing placebo. Thus I believe there should be a way to explain the phenomenon in technical terms, if just digging deep enough into theory and ultra-precise measurement.
It appears that the science you wish to discuss is actually neuroscience.
My MG1.6s have four parallel capacitors on the quasi-ribbon tweeter: 10uF, 6.8uF, 5.1uF, and 0.1uF. The 0.1uF capacitor ensures that the impedance does not rise at ultrasonic frequencies. That may have been problematic for some amplifiers.
The crossover also has a 4A fuse that can only be blown by an amplifier going unstable. I suspect that Magnepan may have had some experience with unstable amplifiers.
Ed
The crossover also has a 4A fuse that can only be blown by an amplifier going unstable. I suspect that Magnepan may have had some experience with unstable amplifiers.
Ed
That too. If we discuss audio science, neuroscience is included. From this thread though, I hope primarily for insights into bypass capacitors in circuits.
It is an interesting and important notion that it's the amplifier that could be drawing the benefit from bypass caps. Would hope to know more, what interactions could manifest into the audible range.
Well, now that I've learned a bit, it seems that bypassing should be done in a cascaded manner, with the valley-shaped impedance curves of each capacitor known. Getting the impedance lower towards RF would go smoothly and without kinks if the resonant frequencies of each successive bypass (of a bypass...) were aligned for the most linear overall impedance versus frequency. This could go on and on, and end in a fixed vacuum capacitor even.
Is the consensus then, that the audible benefit of having this lower impedance reach well into RF, comes from interaction with amplifiers?
Could it also relate to having an easier path for the RF? We don't want it going through the driver because of IM, so making its path easier to the driver seems contradictory, but could we imagine that it's also an easier path for it to exit, and not linger at the driver? I know, I know, RF doesn't exactly "linger", I'm just speculating.
Is the consensus then, that the audible benefit of having this lower impedance reach well into RF, comes from interaction with amplifiers?
Could it also relate to having an easier path for the RF? We don't want it going through the driver because of IM, so making its path easier to the driver seems contradictory, but could we imagine that it's also an easier path for it to exit, and not linger at the driver? I know, I know, RF doesn't exactly "linger", I'm just speculating.
The 0.1uF capacitor is just to keep the impedance low at radio frequencies. This is likely a "don't care" for 99% of amplifiers.
Ed
Ed
The ceramic 100nF decoupling caps on opamps are to stop them oscillating at RF - if this happens the performance drops to abysmal levels typically (can be high distortion, or noise correlated with the audio signal, definitely a bad thing). Different opamps require different amounts of high-speed decoupling - some require the decoupling from both rails to ground, some are happy with just one cap between the two rails, some like more than 100nF for stability especially if the power supply is remote. 100nF is a generic value that's generally used for any high-speed decoupling as it tends to work for most situations (datasheet may be more specific, if the datasheet says nothing about decoupling then assume 100nF).
High speed ceramic decoupling like this is not to do with noise on the rails, it just has to be enough to stabilize the amplifier. The job of quieting the noise on the rails is what bulk (electrolytic) capacitance does - this has effect at audio frequencies where the noise would be a problem.
Bulk capacitance also has a role in high-gain amplifiers - it prevents unintended feedback paths via the supply rails that could cause the whole circuit to oscillate. Often RC filtering is used, not just decoupling, for this purpose. General purpose microphone preamps are a good example, some have 70dB or more total gain at highest sensitivity.
So its not to do with audible improvement, its whether the circuit is stable or not (if not, its a fault that needs correcting). I've seen opamps with inadequate decoupling suddenly switch between <0.001% distortion to 0.1% distortion on higher signal levels for instance - I added more decoupling and this completely stopped. (This was also on a breadboard which is often less stable than a PCB).
With high speed decoupling one issue that may not be obvious is just how high impedance supply rails are without it - stray inductance becomes dominant over trace resistance at even modest RF frequencies of a few MHz, without placing decoupling caps next to each chip the supply impedance would be high at RF, even with thick PCB traces. Try to place the caps within 5mm of the supply pins ideally, especially for faster opamps (higher GBW product).
[ in retrospect this may be talking about passive crossovers - no need for RF decoupling for passive crossovers, no active devices are present that might oscillate ]
High speed ceramic decoupling like this is not to do with noise on the rails, it just has to be enough to stabilize the amplifier. The job of quieting the noise on the rails is what bulk (electrolytic) capacitance does - this has effect at audio frequencies where the noise would be a problem.
Bulk capacitance also has a role in high-gain amplifiers - it prevents unintended feedback paths via the supply rails that could cause the whole circuit to oscillate. Often RC filtering is used, not just decoupling, for this purpose. General purpose microphone preamps are a good example, some have 70dB or more total gain at highest sensitivity.
So its not to do with audible improvement, its whether the circuit is stable or not (if not, its a fault that needs correcting). I've seen opamps with inadequate decoupling suddenly switch between <0.001% distortion to 0.1% distortion on higher signal levels for instance - I added more decoupling and this completely stopped. (This was also on a breadboard which is often less stable than a PCB).
With high speed decoupling one issue that may not be obvious is just how high impedance supply rails are without it - stray inductance becomes dominant over trace resistance at even modest RF frequencies of a few MHz, without placing decoupling caps next to each chip the supply impedance would be high at RF, even with thick PCB traces. Try to place the caps within 5mm of the supply pins ideally, especially for faster opamps (higher GBW product).
[ in retrospect this may be talking about passive crossovers - no need for RF decoupling for passive crossovers, no active devices are present that might oscillate ]
Last edited:
This shouldn't be difficult. How long are the speaker cables and what type of wire are they? What amplifier? And this is a passive crossover in a speaker, right?
Also, can you post a schematic of the crossover and bypass caps? Let's see what you have going there, so we can get a better idea 🙂
Also, can you post a schematic of the crossover and bypass caps? Let's see what you have going there, so we can get a better idea 🙂
If there is incidental resistance configured with the bypass cap, it could form a low pass filter.
Not decoupling. However the speakers are at one end of transmission line, at the other end of which resides the power amp. The reflected impedance from the speaker can affect some power amps, depending on the particulars.
First there should be some sonic differences which I doubt unless proven otherwise, and second by being so small, relative influence would be minimized by 100X or whatever the capacitance ratio would've, so ....
We are worrying about the weight of mosquito and even worse, 1/hundredth of a mosquito.
Credible seeming is a very elastic and undefined label.
"Saying* they hear something, same thing.
If at least all (or at least 60/70% ) agreed then perhaps we should investigate it further, but opinions are all over the map, so ....
If anything, a small plastic cap will be way more linear than a larger one, specially if the large one is electrolytic.
And if both plastic, both would be in theme league.
If understood correctly a larger cap would start rolling of high frequency.
The intent would seem impedance compensation
So audible = nothing because tweeter cant reproduce frequency
or amplifier or your brain or ears.
If audible , likely assume amplifier is junk or oscillating
Some metal domes have strange peaks above 20 K
so could roll off response. Likely, Female can complain
Males dont hear. Depends on source recording.
Not much above 8 or 12k , but " clarity" may be noticeable
with good tweeter.
Far as older recordings, the extra "detail" is microphone distortion
being more noticeable. If your familiar with old Jazz recordings.
Dont care to analyze capacitor magic to much.
Since anything actually audible is mainly in recording.
And microphones not doing much over certain range.
The intent would seem impedance compensation
So audible = nothing because tweeter cant reproduce frequency
or amplifier or your brain or ears.
If audible , likely assume amplifier is junk or oscillating
Some metal domes have strange peaks above 20 K
so could roll off response. Likely, Female can complain
Males dont hear. Depends on source recording.
Not much above 8 or 12k , but " clarity" may be noticeable
with good tweeter.
Far as older recordings, the extra "detail" is microphone distortion
being more noticeable. If your familiar with old Jazz recordings.
Dont care to analyze capacitor magic to much.
Since anything actually audible is mainly in recording.
And microphones not doing much over certain range.
Try this old trick: Put a 100R non-inductive resistor across the speaker terminals. Why? Twisted pair wire usually has a characteristic impedance of around 100R - 120R. The idea is to terminate the transmission line that is the speaker cable to dampen RF reflections that can affect operational characteristics and or stability of some amplifiers. Despite the fact that the causation is at RF frequencies, there are mechanisms by which it can cause audible effects.
hmm.. the transmission line impedance can be varying all over the place, what regards real power cabling for speakers.. I would say from some ohm to hundreds of ohm, depending on exact cable geometry..?
True. The 100R non-inductive resistor doesn't always do anything. But sometimes it does the trick. In any case it may look better than an open circuit at up at 1MHz - 2MHz. PMA has described the risks of an open if the cable is long enough.
A blinded hearing test would need to be done to see if an audible difference is discernable.
The inductance and capacitance of several feet of speaker cable will likely negate the effect of any small value capacitor in a speaker crossover. Of course a simple measurement with a microphone will show any audible difference. A nice LTSpice model can be used to predict the expected result. In school we were not allowed to start the lab experiments and measurements until we had completed calculating the expected results and documenting them in the lab notebook. Audiophiles don't give a rip about science, so they will buy the platinum plated gold foil capacitors with snake oil and diamond coating, and create a sapphire crystal window on the back of the speaker to display that capacitor as it is bolted into the crossover with 2/0 200 Amp terminal block connections. They will write heart melting poetry about the change in the liquid highs and the lifted veils while having a lot of fun doing it too. The folks here for HiFi will make a calculation and maybe a measurement to verify the calculated result and not bother with it if the effect is too small to matter. These are two different hobbies, each can be fun.
Last edited: